1 / 76

CS514: Intermediate Course in Operating Systems

CS514: Intermediate Course in Operating Systems. Professor Ken Birman Vivek Vishnumurthy: TA. Scalability. Today we’ll focus on how things scale Basically: look at a property that matters Make something “bigger” Like the network, the number of groups, the number of members, the data rate

kalil
Download Presentation

CS514: Intermediate Course in Operating Systems

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. CS514: Intermediate Course in Operating Systems Professor Ken BirmanVivek Vishnumurthy: TA

  2. Scalability • Today we’ll focus on how things scale • Basically: look at a property that matters • Make something “bigger” • Like the network, the number of groups, the number of members, the data rate • Then measure the property and see impact • Often we can “hope” that no slowdown would occur. But what really happens?

  3. Stock Exchange Problem: Sometimes, someone is slow… Most members are healthy…. … but one is slow i.e. something is contending with the red process,delaying its handling of incoming messages…

  4. 32 96 With a slow receiver, throughput collapses as the system scales up Virtually synchronous Ensemble multicast protocols 250 group size: 32 group size: 64 group size: 96 200 150 average throughput on nonperturbed members 100 50 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 perturb rate

  5. Why does this happen? • Superficially, because data for the slow process piles up in the sender’s buffer, causing flow control to kick in (prematurely) • But why does the problem grow worse as a function of group size, with just one “red” process? • Small perturbations happen all the time

  6. Broad picture? • Virtual synchrony works well under bursty loads • And it scales to fairly large systems (SWX uses a hierarchy to reach ~500 users) • From what we’ve seen so far, this is about as good as it gets for reliability • Recall that stronger reliability models like Paxos are costly and scale far worse • Desired: steady throughput under heavy load and stress

  7. Protocols famous for scalability • Scalable reliable multicast (SRM) • Reliable Multicast Transport Protocol (RMTP) • On-Tree Efficient Recovery using Subcasting (OTERS) • Several others: TMP, MFTP, MFTP/EC... • But when stability is tested under stress, every one of these protocols collapses just like virtual synchrony!

  8. Example: Scalable Reliable Multicast (SRM) • Originated in work on Wb and Mbone • Idea is to do “local repair” if messages are lost, various optimizations keep load low and repair costs localized • Wildly popular for internet “push,” seen as solution for Internet radio and TV • But receiver-driven reliability model lacks “strong” reliability guarantees

  9. Local Repair Concept

  10. Local Repair Concept

  11. Local Repair Concept lost

  12. Local Repair Concept Receipt of subsequent packet triggers NACK for missing packet NACK NACK X NACK

  13. Local Repair Concept Retransmit NACK NACK X X NACK X Receive useless NAK, duplicate repair

  14. Local Repair Concept X X X NACK X NACK X X Receive useless NAK, duplicate repair X

  15. Local Repair Concept X NACK X X NACK Receive useless NAK, duplicate repair NACK X

  16. Local Repair Concept X X X Receive useless NAK, duplicate repair

  17. Limitations? • SRM runs in application, not router, hence IP multicast of nack’s and retransmissions tend to reach many or all processes • Lacking knowledge of who should receive each message, SRM has no simple way to know when a message can be garbage collected at the application layer • Probabilistic rules to suppress duplicates

  18. In practice? • As the system grows large the “probabilistic suppression” fails • More and more NAKs are sent in duplicate • And more and more duplicate data message are sent as multiple receivers respond to the same NAK • Why does this happen?

  19. Visualizing how SRM collapses • Think of sender as the hub of a wheel • Messages depart in all directions • Loss can occur at many places “out there” and they could be far apart… • Hence NAK suppression won’t work • Causing multiple NAKS • And the same reasoning explains why any one NAK is likely to trigger multiple retransmissions! • Experiments have confirmed that SRM overheads soar with deployment size • Every message triggers many NAKs and many retransmissions until the network finally melts down

  20. Dilemma confronting developers • Application is extremely critical: stock market, air traffic control, medical system • Hence need a strong model, guarantees • But these applications often have a soft-realtime subsystem • Steady data generation • May need to deliver over a large scale

  21. Today introduce a new design pt. • Bimodal multicast (pbcast) is reliable in a sense that can be formalized, at least for some networks • Generalization for larger class of networks should be possible but maybe not easy • Protocol is also very stable under steady load even if 25% of processes are perturbed • Scalable in much the same way as SRM

  22. Environment • Will assume that most links have known throughput and loss properties • Also assume that most processes are responsive to messages in bounded time • But can tolerate some flakey links and some crashed or slow processes.

  23. Start by using unreliable multicast to rapidly distribute the message. But some messages may not get through, and some processes may be faulty. So initial state involves partial distribution of multicast(s)

  24. Periodically (e.g. every 100ms) each process sends a digest describing its state to some randomly selected group member. The digest identifies messages. It doesn’t include them.

  25. Recipient checks the gossip digest against its own history and solicits a copy of any missing message from the process that sent the gossip

  26. Processes respond to solicitations received during a round of gossip by retransmitting the requested message. The round lasts much longer than a typical RPC time.

  27. Delivery? Garbage Collection? • Deliver a message when it is in FIFO order • Garbage collect a message when you believe that no “healthy” process could still need a copy (we used to wait 10 rounds, but now are using gossip to detect this condition) • Match parameters to intended environment

  28. Need to bound costs • Worries: • Someone could fall behind and never catch up, endlessly loading everyone else • What if some process has lots of stuff others want and they bombard him with requests? • What about scalability in buffering and in list of members of the system, or costs of updating that list?

  29. Optimizations • Request retransmissions most recent multicast first • Idea is to “catch up quickly” leaving at most one gap in the retrieved sequence

  30. Optimizations • Participants bound the amount of data they will retransmit during any given round of gossip. If too much is solicited they ignore the excess requests

  31. Optimizations • Label each gossip message with senders gossip round number • Ignore solicitations that have expired round number, reasoning that they arrived very late hence are probably no longer correct

  32. Optimizations • Don’t retransmit same message twice in a row to any given destination (the copy may still be in transit hence request may be redundant)

  33. Optimizations • Use IP multicast when retransmitting a message if several processes lack a copy • For example, if solicited twice • Also, if a retransmission is received from “far away” • Tradeoff: excess messages versus low latency • Use regional TTL to restrict multicast scope

  34. Scalability • Protocol is scalable except for its use of the membership of the full process group • Updates could be costly • Size of list could be costly • In large groups, would also prefer not to gossip over long high-latency links

  35. Can extend pbcast to solve both • Could use IP multicast to send initial message. (Right now, we have a tree-structured alternative, but to use it, need to know the membership) • Tell each process only about some subset k of the processes, k << N • Keeps costs constant.

  36. Router overload problem • Random gossip can overload a central router • Yet information flowing through this router is of diminishing quality as rate of gossip rises • Insight: constant rate of gossip is achievable and adequate

  37. Limited bandwidth (1500Kbits) Bandwidth of other links:10Mbits Receiver Sender High noise rate Receiver Sender Latency to delivery (ms) Load on WAN link (msgs/sec)

  38. Hierarchical Gossip • Weight gossip so that probability of gossip to a remote cluster is smaller • Can adjust weight to have constant load on router • Now propagation delays rise… but just increase rate of gossip to compensate

  39. Remainder of talk • Show results of formal analysis • We developed a model (won’t do the math here -- nothing very fancy) • Used model to solve for expected reliability • Then show more experimental data • Real question: what would pbcast “do” in the Internet? Our experience: it works!

  40. Idea behind analysis • Can use the mathematics of epidemic theory to predict reliability of the protocol • Assume an initial state • Now look at result of running B rounds of gossip: converges exponentially quickly towards atomic delivery

  41. Either sender fails… … or data gets through w.h.p.

  42. Failure analysis • Suppose someone tells me what they hope to “avoid” • Model as a predicate on final system state • Can compute the probability that pbcast would terminate in that state, again from the model

  43. Two predicates • Predicate I: A faulty outcome is one where more than 10% but less than 90% of the processes get the multicast … Think of a probabilistic Byzantine General’s problem: a disaster if many but not most troops attack

  44. Two predicates • Predicate II: A faulty outcome is one where roughly half get the multicast and failures might “conceal” true outcome … this would make sense if using pbcast to distribute quorum-style updates to replicated data. The costly hence undesired outcome is the one where we need to rollback because outcome is “uncertain”

  45. Two predicates • Predicate I: More than 10% but less than 90% of the processes get the multicast • Predicate II: Roughly half get the multicast but crash failures might “conceal” outcome • Easy to add your own predicate. Our methodology supports any predicate over final system state

  46. Figure 5: Graphs of analytical results

More Related